.libPaths("~/renv/renv/library/R-4.5/aarch64-apple-darwin20/")
library(pmplots)
library(pmtables)
library(yspec)
library(dplyr)
library(here)
library(ggplot2)
library(data.table)
library(lastdose)
library(yspec)
library(patchwork)Exploratory Data Analysis - nmdat1.csv
full <- fread(here("model1/data/nmdat1.csv"), na.strings = ".")The data spec will help us format tables and figures
spec <- ys_load(here("wk02/nmdat1.yaml"))
table <- ys_get_short_unit(spec)
full <- ys_factors(full, spec, GENO, SEX, DOSE)Calculate time after last dose
full <- lastdose(full)
head(full) C NUM ID TIME AMT CMT EVID DV BLQ LLOQ AGE SEX
<lgcl> <int> <int> <num> <int> <int> <int> <num> <int> <int> <int> <fctr>
1: NA 1 1 0.00 50 1 1 0.00 0 50 13 Male
2: NA 2 1 0.08 0 2 0 228.98 0 50 13 Male
3: NA 3 1 0.27 0 2 0 565.98 0 50 13 Male
4: NA 4 1 0.49 0 2 0 791.70 0 50 13 Male
5: NA 5 1 0.91 0 2 0 964.83 0 50 13 Male
6: NA 6 1 1.58 0 2 0 1040.88 0 50 13 Male
ALB SCR WT HT GENO BSA EGFR DOSE SEX_v GENO_v
<num> <num> <num> <num> <fctr> <num> <num> <fctr> <int> <int>
1: 4.3 0.8 55.9 168.2 Normal metabolizer 1.63 143.18 50 mg 1 0
2: 4.3 0.8 55.9 168.2 Normal metabolizer 1.63 143.18 50 mg 1 0
3: 4.3 0.8 55.9 168.2 Normal metabolizer 1.63 143.18 50 mg 1 0
4: 4.3 0.8 55.9 168.2 Normal metabolizer 1.63 143.18 50 mg 1 0
5: 4.3 0.8 55.9 168.2 Normal metabolizer 1.63 143.18 50 mg 1 0
6: 4.3 0.8 55.9 168.2 Normal metabolizer 1.63 143.18 50 mg 1 0
DOSE_v TAD LDOS OCC
<int> <num> <num> <num>
1: 50 0.00 50 1
2: 50 0.08 50 1
3: 50 0.27 50 1
4: 50 0.49 50 1
5: 50 0.91 50 1
6: 50 1.58 50 1count(full, OCC) OCC n
<num> <int>
1: 1 34351 Get Started
1.1 Basic inventory of the data
- Include all observation records (
EVID==0) and anyBLQrecords. - I’m stratifying (or splitting) the data by all the categorical covariates I see in the data.
- We also get the inventory for all the data.
data <- filter(full, EVID==0)
pt_inventory_long(
data,
cols = c("DOSE", "GENO", "SEX"),
table = table
) %>% st_as_image()
1.2 Make a simple plot
ggplot(data, aes(TIME, DV)) + geom_point()
1.3 Customize the plot
- I like to update the x-axis tick labels to be more intuitive
ggplot(data, aes(TIME, DV)) + geom_point() +
scale_x_continuous(breaks = seq(0,52,4))
- And we frequently would rater see this on log scale
#| fig-width: 7
#| fig-height: 5
ggplot(data, aes(TIME, DV)) +
geom_point() +
scale_x_continuous(breaks = seq(0,52,4)) +
scale_y_log10()
- I like to have a tick label above the highest observation
ggplot(data, aes(TIME, DV)) +
geom_point() +
scale_x_continuous(breaks = seq(0,52,4)) +
scale_y_log10(breaks = logbr3(), limits = c(NA, 10000))
- Let’s drop the BLQ observations
obs <- filter(data, BLQ==0)
ggplot(obs, aes(TIME, DV)) +
geom_point() +
scale_x_continuous(breaks = seq(0,52,4)) +
scale_y_log10(breaks = logbr3(), limits = c(30, 10000))
- Let’s give this proper labels
ggplot(obs, aes(TIME, DV)) +
geom_point() +
scale_x_continuous(breaks = seq(0,52,4)) +
scale_y_log10(breaks = logbr3(), limits = c(30, 10000)) +
labs(x = "Time (h)", y = "Demothizine concentration (ng/mL)")
1.4 Do you want to know more about absorption?
Make a view specific for that question
abs <- filter(obs, TIME < 8)
ggplot(abs, aes(TIME, DV)) +
geom_point() +
scale_x_continuous(breaks = seq(0,8,2)) +
scale_y_log10(breaks = logbr3(), limits = c(30, 10000)) +
labs(x = "Time (h)", y = "Demothizine concentration (ng/mL)")
2 Dig deeper into the covariates
- First, get a data set with baseline covariates
id <-
obs %>%
slice(1, .by = ID) %>%
mutate(across(c(SEX, DOSE, GENO), ~ factor(.x)))2.1 Covariates - tabular summary
pt_demographics(
id,
cols_cat = c("SEX", "GENO"),
cols_cont = c("ALB", "WT", "AGE", "EGFR"),
span = c("Demothizine dose" = "DOSE"),
table = table
) %>% st_as_image()
pt_demographics(
id,
cols_cat = c("SEX", "DOSE"),
cols_cont = c("ALB", "WT", "AGE", "EGFR"),
span = c(Genotype = "GENO"),
table = table
) %>% st_as_image()
2.2 Covariates - graphical summary
pl <- list_plot_x(
id,
x = c("SEX", "DOSE", "GENO"),
y = c("ALB", "WT", "BSA", "EGFR"),
.fun = cont_cat
)
pl <- lapply(pl, pm_grid, ncol = 2)
pl$SEX
$DOSE
$GENO
- What’s the difference with these plots?
pl <- list_plot_y(
id,
x = c("SEX", "DOSE", "GENO"),
y = c("ALB", "WT", "BSA", "EGFR"),
.fun = cont_cat
)
pl <- lapply(pl, pm_grid, ncol = 2)
pl$ALB
$WT
$BSA
$EGFR
2.3 Covariates - correlations
y <-
spec %>%
ys_namespace("plot") %>%
axis_col_labs(vars(ALB, WT, AGE, EGFR))
pairs_plot(id, y = y)
3 How does PK differ across different covariates?
3.1 By dose
ggplot(obs, aes(TIME, DV)) +
geom_point() +
scale_x_continuous(breaks = seq(0,52,4)) +
scale_y_log10(breaks = logbr3(), limits = c(30, 10000)) +
labs(x = "Time (h)", y = "Demothizine concentration (ng/mL)") +
facet_grid(~DOSE)
- I wouldn’t rely on color to separate the data
ggplot(obs, aes(TIME, DV, col = factor(DOSE))) +
geom_point() +
scale_x_continuous(breaks = seq(0,52,4)) +
scale_y_log10(breaks = logbr3(), limits = c(30, 10000)) +
labs(x = "Time (h)", y = "Demothizine concentration (ng/mL)")
3.2 By genotype
ggplot(obs, aes(TIME, DV)) +
geom_point() +
scale_x_continuous(breaks = seq(0,52,4)) +
scale_y_log10(breaks = logbr3(), limits = c(30, 10000)) +
labs(x = "Time (h)", y = "Demothizine concentration (ng/mL)") +
facet_grid(DOSE~GENO)
ggplot(obs, aes(TIME, DV)) +
geom_point() +
scale_x_continuous(breaks = seq(0,52,4)) +
scale_y_log10(breaks = logbr3(), limits = c(30, 10000)) +
geom_hline(yintercept = c(300, 1000, 3000), lty = 2) +
labs(x = "Time (h)", y = "Demothizine concentration (ng/mL)") +
facet_grid(DOSE~GENO)
3.3 Try boxplots at a nominal time point for categorical covariates
cmin <- filter(data, TIME > 20 & TIME < 25)
ggplot(cmin) +
geom_boxplot(aes(x = factor(GENO), y = DV)) +
scale_y_log10() +
facet_wrap(~DOSE) +
labs(y = "Demothizine concentration (ng/mL)", x = "Genotype")
3.3.1 Do we see the same by SEX?
ggplot(cmin) +
geom_boxplot(aes(x = factor(SEX), y = DV)) +
scale_y_log10() +
facet_wrap(~DOSE) +
labs(y = "Demothizine concentration (ng/mL)", x = "Sex")
3.4 What about continuous?
ggplot(cmin, aes(x = EGFR, y = DV)) +
geom_point() +
geom_smooth()
xx <- axis_col_labs(spec, vars(EGFR, ALB, WT, AGE))
p <- cont_cont(
cmin,
x = xx,
y = "DV//Demothizine Cmin (ng/mL)",
ys = list(limits = c(0, NA))
)
p <- lapply(p, layer_s)
pm_grid(p) + plot_layout(axes = "collect")